Correction of temporal misregistration artifacts in jet flow by conventional phase‐contrast MRI

Abstract

Purpose: To show that accuracy of jet flow representation by magnetic resonance (MR) phase-contrast (PC) velocityencoded (VE) cine imaging is dominated by error terms resulting from the temporal distribution of data, and to present a generally applicable data interpolation-based approach to correct for this phenomenon. Materials and Methods:Phase-contrast data were acquired in a stenotic orifice flow phantom using a physiologic pulsatile flow waveform. A temporally registered scan, acquired without data segmentation or interleaving was obtained (17 minutes) and taken as the reference (REF). Conventional PC data sets were acquired using segmentation and data interleaving. An enhanced temporal registration (ETR) algorithm was applied to the acquired data to temporally interpolate component sets and output data at matching time points, thereby reducing temporal dispersion. Results:Compared to the REF data, conventionally processed PC data consistently overestimated peak velocities in laminar jet flow regions (127% Ϯ 28%) and exhibited relatively weak correlations (r ϭ 0.67 Ϯ 0.23). The ETRprocessed data better represented peak velocities (101% Ϯ 13%, P Ͻ 0.001) and correlated more closely with the REF data (r ϭ 0.94 Ϯ 0.05, P Ͻ 0.001). Conclusion:The temporal distribution of PC data impacts the accuracy of velocity representation in pulsatile jet flow. A temporal registration postprocessing algorithm can minimize loss of accuracy.

Numerical and experimental study of a novel phase contrast magnetic resonance (PC‐MR) imaging technique: Sparse interleaved referencing PC‐MR imaging

“…However, our research group proposed another type of acceleration related error that cannot be corrected by simply interpreting the data at the "gravity" center (17). From our simulated results and experiments, this type of error was shown to be a major contributor to peak velocity inaccuracy in jet flow, seen here in our simulated and experimental data (Figs.…”

“…Doyle et al used a stenotic orifice phantom with physiologic pulsatile flow waveform to investigate the transient acceleration on the accuracy of PC-MR flow measurement. The results show that the Con PC-MR method consistently overestimated peak velocities in laminar jet flow regions by approximately 30% (127% Ϯ 28%) when compared with reference scans (11). A study by Thompson et al shows that, when temporal resolution of the in vivo PC-MR data decreases from 11.2 ms to 56 ms, the peak can be overestimated by 300% at the time points where velocities change sharply (high transient acceleration; 10).…”

“…Further details concerning this type of error are discussed below. Additional errors associated with uncompensated higher order motion components (17) and/or the time delay between the phase encoding and phase readout steps (12) are typically much smaller than the above errors and are not incorporated in our simulations. The major source of random errors comes from white noise contamination (18).…”

“…However, when high segmentation values are used, accuracy in cardiovascular PC-MR is compromised when flow fields with high acceleration terms are present (e.g., valvular regurgitation or stenosis). The loss of accuracy is related to several factors (10), including the relatively poor temporal resolution associated with use of a high number of views per segment (VPS), the data extraction process requiring comparison of velocity encoded (⌽ VE ) and velocity compensated (⌽ VC ) data sets at each time point, making the Con PC-MR technique intrinsically sensitive to temporal accelera-tion (11), and noise amplification brought about by the phase subtraction operation, leading to substantial noise contamination. Doyle et al used a stenotic orifice phantom with physiologic pulsatile flow waveform to investigate the transient acceleration on the accuracy of PC-MR flow measurement.…”

Improved in vivo human carotid artery wall T2⁎ estimation

“…As seen in the carotid artery, even small changes in vessel diameter can cause different tissue or blood signal intensity to replace the vessel wall in images at different echo times. Misregistration may be reduced by the current sequence design using interleaving TEs [20], but even with perfect registration, most of these voxel samples lie along the boundary of the vessel wall and are most likely susceptible to partial volume mixing between surrounding tissue and blood flow. Especially in the case of arterial blood with a much longer T 2 *, this results in a gross overestimation of T 2 * that can dominate the vessel cross section being analyzed.…”